Agents

LiveKit Agents is a Python framework for realtime, programmable participants that run on a server rather than in the browser, aimed at teams building voice assistants, support lines and tutoring or interview bots. The agents it produces are multimodal: they listen, speak, see and hold a conversation while audio and video stream over WebRTC. Nothing is locked to one vendor, so STT, LLM, TTS and realtime speech APIs can be mixed to suit the use case.

Job scheduling and dispatch APIs ship with the framework, which means connecting a user to a free agent is not glue code someone has to write. Sitting on LiveKit's telephony stack lets the same agent place and answer phone calls, and RPC and data APIs carry structured state between the agent and client apps.